Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Emergent s-wave interactions between identical fermions

in quasi-one-dimensional geometries
Kenneth G. Jackson,1 Colin J. Dale,1 Jeff Maki,2 Kevin G. S. Xie,1 Ben A.
Olsen,3 Denise J. M. Ahmed-Braun,4 Shizhong Zhang,2 and Joseph H. Thywissen1
1
Department of Physics and CQIQC, University of Toronto, Toronto, Ontario M5S 1A7, Canada
2
Department of Physics and HKU-UCAS Joint Institute for Theoretical and Computational Physics at Hong Kong,
Guangdong-Hong Kong Joint Laboratory of Quantum Matter,
The University of Hong Kong, Hong Kong, China
3
Yale-NUS College, Singapore 138527, Singapore
arXiv:2206.10415v1 [cond-mat.quant-gas] 21 Jun 2022

4
Department of Physics, Eindhoven University of Technology, The Netherlands
(Dated: June 22, 2022)
One-dimensional models of spin-polarized Fermi gases with tunable zero-range interactions have
been the subject of numerous theoretical investigations, but are as yet unrealized in ultracold sys-
tems. Here we study the two-body correlation strength of spin-polarized fermionic potassium (40 K)
in the quasi-one-dimensional (q1D) regime, using radio-frequency spectroscopy. The strength and
spatial symmetry of interactions are tuned by a nearby p-wave Feshbach resonance and by con-
finement anisotropy. Surprisingly, we find a scattering channel that has even particle-exchange
parity along the q1D axis. These emergent s-wave collisions are enabled by orbital singlet wave
functions in the transverse directions, which also confer high-momentum components to low-energy
q1D collisions. We measure both the q1D odd- and even-wave “contact” parameters for the first
time, and compare them to theoretical predictions of one-dimensional many-body models. We find
an even-wave contact that is comparable to the theoretical unitary value, whereas the maximum
observed odd-wave contact is two orders of magnitude smaller than its unitary limit. The discovery
of emergent s-wave interactions reveals a new route to studying strongly correlated ensembles of
spin-polarized fermions with zero-range interactions.

I. INTRODUCTION where V and R3D are the p-wave scattering volume and
effective range, respectively, and ~k is the relative mo-
Broad interest in the many-body physics of ultracold mentum. Near a p-wave Feshbach resonance, V is widely
atoms arises from regimes of universality, in which phe- tunable, but the Wigner bound [15–17] constrains the
nomena can be understood with models independent of maximum value of R3D to be on the order of the van der
the particular element used in experiments. Examples Waals length, which is small compared to typical k −1 in
include Hubbard models, two-dimensional bosons near the ultracold regime. Thus (kR3D )−1 remains relevant
their non-interacting fixed point, and three-dimensional for p-wave resonant scattering.
(3D) fermionic spin mixtures near an s-wave scattering Strong one-dimensional (1D) confinement has been
resonance. In these universal systems, one can often iden- proposed to both suppress the inelastic losses close to p-
tify a solitary relevant parameter, for instance kF as for wave resonances [18, 19] and also to achieve a zero-range
3D Fermi gases, where kF is the Fermi momentum and regime [20, 21]. The 1D odd-wave scattering amplitude
as the s-wave scattering length. The tunability of as in fo is given by
certain systems, via Feshbach resonances [1], has allowed − fo−1 = 1 − i(k`o )−1 − iro k + O(k 3 ) , (2)
a full exploration of universal s-wave phenomena [2, 3].
The leading partial wave for spin-polarized ultracold where `o and ro are the 1D odd-wave scattering length
Fermi gases is p-wave, which differs from the s-wave col- and effective range, respectively. The 1D unitary limit
lisional channel in several ways. First, the existence of a fo → −1 is achieved when `−1 o  k  ro−1 , in con-
centrifugal barrier limits the coupling strength between trast to the 3D p-wave interactions. In practice, how-
short-range dimers and scattering states, resulting in an ever, accessing k  ro−1 is difficult because of the na-
energetically narrow resonance. Also, near a Feshbach ture of the confinement-induced resonance, and no exper-
resonance, scattering is accompanied by strong three- iment to date has been able to enter the universal odd-
body loss [4–9], due to an enhanced amplitude for the wave regime. In addition, investigations with quasi-one-
particles to be at short inter-nuclear distances behind dimensional (q1D) ensembles of 6 Li have found strong
the centrifugal barrier. Finally, it has been realized both p-wave losses near resonance [22, 23].
experimentally [8, 10, 11] and theoretically [12–15] that In this work, we explore the universal and q1D physics
to characterize the low-energy scattering properties of a of a spin-polarized Fermi gas of 40 K. As depicted in
p-wave gas correctly, it is crucial to include the effective Fig. 1(a), atoms in an array of q1D tubes are probed
range parameter in the scattering amplitude: with radio-frequency (rf) spectroscopy. The transfer rate
between two internal spin states is related to the short-
− (kf3D )−1 = i + (V k 3 )−1 + (kR3D )−1 + O(k) , (1) range correlations between atoms through universal con-
2

(a) (b) Ground-band loading (c) Multi-band loading


s
ire λL /2 140
rf w
y –1
B0 (kF lo) = –10
y 1.0
120 B0x

Lattice depth (ER)

Transfer Γ (10 )
–4
3
z x

=
100

z
e,
a
B

k
F

~
80 0.5
g
60

40
0.0
20
198 199 200 201 202 203 198 199 200 201 202 203
Magnetic field (G) Magnetic field (G)

FIG. 1. Summary of results. (a) An array of q1D traps aligned in the vertical (y) direction is created by two orthogonal standing
waves of light at wavelength λL . A static magnetic field B is oriented along z. Radio-frequency fields are generated by a pair of
proximate microfabricated conductors. (b,c) The normalized rf transfer rate Γ, e which is a measure of correlation strength (see
Sec. III), is indicated by color across a range of magnetic fields and optical lattice depths. The depth of the confining optical
lattice, VL , is given in units of the recoil energy ER = h2 /2mλ2L . The first peak in Γ e lies between the odd-wave resonance
y
(indicated by a solid black line at B0 , where `o = 0 as discussed in Sec. II) and (kF `o )−1 = −10 (indicated by a black dashed
−1

line). Panel (c): Correlations beyond the odd-wave regime are found in ensembles with population in transverse excited bands.
We show that these additional correlations are due to effective s-wave interactions along the q1D direction. The structure of
Γ(B,
e VL ) and is further explored in Figs. 5, 6, and 8.

tact relations [24–40]. We observe the q1D odd-wave con- In Sec. II, we explain how low-energy collisions that
tact parameter anticipated by Refs. [41–46] for the first are exchange-even along the q1D direction are enabled
time, and compare it to theoretical predictions. These by atom pairs sharing one quantum of transverse orbital
correlations are found for a range of magnetic fields and excitation with singlet character. We derive the scat-
confinement strengths that span −10 . (kF `o )−1 . 0, as tering T -matrix for both odd- and even-wave channels
shown in Fig. 1(b). We show that the maximum contact on equal footing. In Sec. III, we explain our experimen-
parameter is displaced from `−1
o = 0, and remains two or- tal protocol, find the rf spectral signatures of each q1D
ders of magnitude below its expected unitary value, due parity, and delineate conditions under which the rf spin-
to losses similar in strength and origin to those seen in flip rate is a measure of the contact parameter. These
6
Li [22, 23]. relations generalize prior treatments of low-dimensional
Surprisingly, we discover conditions under which scattering to the case in which multiple transverse bands
fermions interact via wave functions with even particle- are relevant. In Sec. IV, we measure the q1D odd-wave
exchange symmetry along the q1D degree of freedom. correlation strength, and show that the contact parame-
These q1D even-wave collisions manifest as correlations ter and atom loss rate are governed by the dimensionless
well beyond the q1D odd-wave parameter range, as coupling strength 1/kF `o . In Sec. V, we present three
shown in Fig. 1(c), and can be made prominent by chang- investigations of the nature of the emergent s-wave inter-
ing the loading procedure for the confined array. In the actions. We find that the dimensionless even-wave con-
low-energy limit, the 1D even-wave scattering amplitude tact parameter per atom is comparable to the expected
is unitary value when kF ae approaches zero, and discuss
why the inferred interaction strength is at least an or-
− fe−1 = 1 + iae k − ive k 3 + O(k 5 ) , (3) der of magnitude stronger than those produced by the
where ae and ve are the 1D even-wave scattering length odd-wave interactions. We conclude in Sec. VI with a
and effective range, respectively. Unlike the odd-wave summary and discussion of future prospects.
case, both the resonant kae  1 and zero-range k 3 ve 
1 limits are achieved at low energy, with no necessary
hierarchy between ae and ve ; this enables q1D even-wave
scattering to enter the universal (zero-range) regime at II. P-WAVE SCATTERING IN A
QUASI-ONE-DIMENSIONAL OPTICAL TRAP
low energy. In fact, the form of Eq. (3) is identical to
the 1D limit of s-wave collisions in spin-half Fermi gases
[47–49]. We emphasize that true s-wave collisions are We consider a Fermi gas of identical (single-
forbidden for spin-polarized fermions; the q1D even-wave component) atoms confined in a highly anisotropic poten-
interactions observed here are due to underlying 3D p- tial, or a “tube”, in the deep lattice limit where tunnel-
wave interactions. ing between tubes can be neglected. The single-particle
3

Hamiltonian is COM degree of freedom is either in its ground state or


1 2 m 2 2 decoupled, we will focus our discussion on the relative
H0 = p + (ωx x + ωy2 y 2 + ωz2 z 2 ) , (4) motion. Note that ~ has been set equal to one here and
2m 2 for the remainder of this derivation.
where p = (px , py , pz ) is the 3D momentum of the atom, In the center-of-mass frame, setting ωx,z = ω⊥ and
and we have assumed that anharmonic corrections are ωy = 0, we write the two-body Hamiltonian as
small. Quasi-one-dimensional confinement is achieved
when the transverse harmonic frequencies in the x and 1 2 µ
H2b = p + ω 2 ρ2 + U , (6)
z direction are much larger than that in the y direction: 2µ rel 2 ⊥
{ωx , ωz }  ωy . This allows a separation of energy scales,
in which the collision energy (determined by the Fermi where prel = (pρ , pφ , k) is the relative momentum and U
energy EF when at low temperature, T . EF /kB ) is is the interatomic interaction described below, with ρ and
much smaller than the energy quantum of the transverse φ as cylindrical coordinates. We assume (for now) that
motion, namely EF  {~ωx , ~ωz }. For N  1 spin- U has cylindrical symmetry, so that we can delineate the
less fermions in a single tube, the 1D Fermi energy is relative wave functions for this trapping geometry by the
EF = N ~ωy , so that the condition for q1D collisions is principal quantum number in the radial direction N and
the azimuthal quantum number m. Up to a normaliza-
Nλ  1 , (5) tion factor, the non-interacting relative wave functions
are
where λ ≡ ωy /ω⊥ is the trap aspect ratio, written in the
ρ2
case of isotropic lateral confinement, ωx = ωz ≡ ω⊥ . As −
2a2
eiky e eimφ ρ|m| ρ2
 

discussed in Sec. III, N λ ∼ 0.1 for our typical experi- |m|
ψN ,m,k (ρ, φ, y) = LN , (7)
mental parameters. a⊥
|m|+1 a2⊥
Low-dimensional confinement has previously been used
in experiments to suppress p-wave collisions along the |m|
where pLN (x) is a generalized Laguerre polynomial, and
strongly confined directions [22, 23, 50, 51] for atoms a⊥ = 1/µω⊥ is the radial oscillator length. The ener-
in the lowest-energy states. For such atoms, higher gies of the ψN ,m,k states are EN ,m,k = k 2 /2µ + (2N +
transverse states are inaccessible, and are only relevant |m| + 1)ω⊥ .
as closed channels [52]; whereas atoms in the ground To incorporate the correct momentum dependence of
band scatter with the expected odd-wave symmetry. p-wave scattering, a model interatomic potential U can
The emergent even-wave scattering found in our exper- be written in the following way. Given two arbitrary
iment requires a generalization that includes population two-body states φ and χ, the matrix element of U is
of asymptotic states with orbital excitation. When hy-
bridized with 3D p-wave dimer states, new modalities of hφ|U |χi = g∇φ∗ (0) · ∇χ(0) , (8)
q1D scattering emerge.
evaluated at zero separation between the two fermions,
where g is a coupling constant [52, 58–60]. In the
A. Scattering T-matrices in q1D regime: the q1D regime, the relevant two-particle states are those in
emergence of even-wave channels
Eq. (7), denoted as |N , m, ki, and the matrix elements
of U in this basis are
Scattering in the q1D low-energy limit has been con-
g
sidered previously for even-wave scattering [47, 53–55] of hN 0 , m0 , k 0 |U |N , m, ki =
kk 0 δm,0 δm,m0
both identical bosons and spin mixtures of fermions, and πa2⊥
also for odd-wave scattering [56, 57] of identical fermions. 2g √ √
+ 4 N + 1 N 0 + 1δ|m|,1 δm,m0 . (9)
Here, we re-examine the two-body problem and present πa⊥
a unified treatment for arbitrary energies when multiple
transverse bands are relevant. This allows us to describe Since the relative wave function vanishes as ψN ,m,k ∝
emergent even-wave scattering on equal footing with odd- ρ|m| , this pseudopoential only has non-zero matrix ele-
wave scattering and to understand the results shown in ments for |m| ≤ 1. The structure of Eq. (9) is as follows.
Fig. 1(c). The first term on the r.h.s. corresponds to m = 0, where
For two non-interacting atoms, the two-body the gradient operator acts along the y direction (parallel
Schrödinger equation can be separated into center- to k and k 0 ). This produces the correct form factor for
of-mass (COM) and relative (rel) motion, each described odd-wave scattering in the q1D direction, kk 0 . As a re-
by a Hamiltonian with the same form as Eq. (4) except sult, it inherits the low-momentum behavior of the 3D p-
for the replacement of the mass m with the total mass wave scattering and vanishes at low energy. On the other
2m or the reduced mass µ = m/2. In the limit when hand, the second term corresponds to the gradient oper-
N  1, we can treat collisions along the q1D axis (y) ator acting along the transverse directions, and as such,
semiclassically, defined by the relative momentum k, but requires antisymmetric radial wave functions (m = ±1)
retain the quantization of transverse motion. Since the to give rise to a non-vanishing matrix element. This term
4

is proportional to a constant at low k and k 0 , which is a with respective solutions given by


hallmark of s-wave scattering.  −1
Due to the cylindrical symmetry of U , the interact- 1 (0)
To−1 (Erel ) ≡ T (E rel ) =
ing two-particle problem separates into three indepen- πa2⊥ q1D
dent sectors, each characterized by a definite value of m. 2µ a2⊥ 1
 
2µErel

2

1

For the m = 0 sector, the two-body wave function is + − ζ − , −E0 − iδ (12)
2 6 V R3D a⊥ 2
exchange-symmetric along the transverse direction: see
Eq. (7) under parity transformation φ → φ+π. The over- and
all exchange antisymmetry of the two-body wave function
−1
then requires the axial wave function to be antisymmetric

2 (±1)
under exchange, and thus two particles undergo odd-wave Te−1 (Erel ) ≡ T (E rel ) =
πa4⊥ q1D
scattering along the q1D direction (also called 1D p-wave
2µ a4⊥ 1
    
scattering). For scattering just above the lowest non- 2µErel a⊥ 1
+ + ζ − , −E1 − iδ
interacting band energy Erel = ω⊥ , the interacting wave 2 12 V R3D 2 2
function consists of all |N , m = 0, ki states, starting with
 
a⊥ Erel 1
N = 0, for which both atoms are in the single-particle + ζ , −E1 − iδ . (13)
2 2ω⊥ 2
ground state of the transverse confinement potential. As
we show below [see Eq. (14)], scattering asymptotically Here ζ(s, x) is the Hurwitz zeta function, δ is an
resembles that of a spinless Fermi gas in 1D. The odd- infinitesimally small positive quantity and Em =
wave amplitude for q1D scattering has been shown pre- [Erel − (|m| + 1)ω⊥ ] /2ω⊥ = (ka⊥ /2)2 . We have also
viously [20, 52, 59, 61, 62] to have the true 1D form, i.e., defined the appropriate renormalized q1D scattering T -
Eq. (2). matrices To (odd wave) and Te (even wave) that have
On the other hand, for the m = 1 and m = −1 sectors, a direct correspondence with scattering T -matrices in
the wave function is odd in the transverse directions [see strictly 1D systems.
again Eq.(7) for φ → φ + π] such that the scattering These two-body T -matrices give their respective scat-
wave function is exchange-even along y. For collision tering amplitudes. For an incident state |N , m, ki, the
energies just above the lowest-energy m = 1 band, Erel & resulting scattered wave with quantum numbers N 0 and
2ω⊥ , the interacting wave function is a superposition of m0 in the transverse direction is given by
|N , m = 1, ki states. In the asymptotic region, the N = 
0 wave function resembles a two-component Fermi gas, ψN 0 m0 (y) = δm0 ,m δN 0 ,N eiky
but in which the role of spin is played by the transverse
oscillator states, as further discussed after Eq. (15).  |m|+1 #
√ 0
(m) y − a2 N −Em −iδ |y|
The separation between the even- and odd-wave chan- +fN 0 e ⊥ , (14)
|y|
nels in Eq. (9) suggests a similar structure for the low-
energy T -matrix in the q1D geometry: (m)
where fN 0 is the scattering amplitude into the N 0 radial
1 (0) state,
hN 0 , m0 , k 0 |Tq1D |N , m, ki = kk 0 Tq1D (Erel )δm,0 δm,m0
πa2⊥ (0)
2 √ √ fN 0 = −iµk To (Erel ) (15)
(1)
+ 4 N + 1 N 0 + 1 Tq1D (Erel )δ|m|,1 δm,m0 , (10) √ √
πa⊥ (±1) a⊥ N 0 + 1 N + 1
and fN 0 = −iµ √ 0 Te (Erel ) .
2 N − E1 − iδ
(m)
where Tq1D is the T -matrix in the q1D geometry with
Consider the m = 0 sector and let the initial asymp-
(m)
relative angular momentum m. To evaluate Tq1D , we use totic state be given by |0, 0, ki with E0  1, then from
the relation between the free-space 3D T -matrix, T3D , Eq. (14) we see that for high-energy transverse oscillator
(m)
and that of the q1D geometry, Tq1D , states with N 0 6= 0, the associated evanescent scattering
wave vanishes for |y|  a⊥ . Only the component with
(m) (m) N = 0 and m = 0 survives asymptotically; this state
Tq1D (Erel ) = T3D (Erel ) (11)
h i describes two particles in the single-particle ground state
(m) (m) (m) (m) of the transverse oscillator.
+ T3D (Erel ) Gq1D (Erel ) − G3D (Erel ) Tq1D (Erel ) ,
Now consider the m = 1 and m = −1 sectors. For an
incoming state |N = 0, m = 1, ki, all the N 0 6= 0 com-
(m) (m)
where Gq1D and G3D are the propagators in the q1D ponents of the scattering wave function similarly van-
geometry and 3D free space, respectively, with definite ish when |y|  a⊥ and E1  1, according to Eq. (14).
angular momentum m [63]. The free-space 3D scattering Therefore the asymptotic q1D wave function features a
matrix, T3D , is (2π/µ)f3D . transverse component that is characterized by N = 0,
Equation (11) can be evaluated analytically for both m = 1. In terms of single-particle wave functions, the
the odd-wave (m = 0) and even-wave (m = ±1) sectors state N = 0, m = 1 represents two atoms sharing a
5

(a) yx z Energy (d) q1D odd

Collision energy (kHz)


0x,1z, k 80
rel
z
Ψrel |fo |2 1.0

Ezp + hωz 60

1x,0z, k rel x 40 0.5

Ezp + hωx y nz 20
0.0
0
x 198.0 198.5 199.0 199.5 200.0 200.5 201.0
0x,0z, k rel z, B nx B (G)
Ezp
(e) q1D even Ψrel

Collision energy (kHz)


80
z |fe|2
y
B0
x
B0
z
B0 Magnetic field 60
(b) 1.0 x
0x,1z, k rel 40
0.8
1x,0z, k rel
0.6 y nz 20
0x,0z, k
| f |2

rel
0.4
0
0.2 x 198.0 198.5 199.0 199.5 200.0 200.5 201.0
z, B nx B (G)
0.0
0 20 40 60 80 100
Collision energy (kHz) (f) q1D even Ψrel 80

Collision energy (kHz)


(c) 1.0
z |fe|2
0x,1z, k rel 60
0.8
1x,0z, k rel x
0.6 40
| f |2

0x,0z, k rel
0.4 y nz 20
0.2
0
0.0 x
198.0 198.5 199.0 199.5 200.0 200.5 201.0
198 199 200 201 202 203 z, B nx B (G)
Magnetic field B (G)

FIG. 2. Scattering resonances in q1D. (a) The asymptotic states |nx , nz , kirel have confinement energies of (nx + 1/2)~ωx +
(nz + 1/2)~ωz plus a continuum of ~2 k2 /2µ from motion along y. The three lowest continua, shown here, become strongly
interacting when magnetically tuned to resonance with an appropriate 3D p-wave dimer state (labelled x, y, and z). The
ground-band resonance location (green disc at B0y ), is lower than the resonance location for |1x , 0z , kirel states (blue square at
B0x ), because the x dimer must reach a continuum that is ~ωx higher in energy than the zero-point energy Ezp of the transverse
confinement. The resonance for |0x , 1z , kirel states (red square at B0z ), is further displaced by the dipole-dipole splitting of
the xy and z dimer states, and any anisotropy in confinement for the case ωz 6= ωx . Corrections to this approximate picture
are given by Eqs. (20) and (21). (b) The effective 1D scattering cross section |f |2 at resonance versus q1D collision energy.
The broader resonance for asymptotic states with nx = 1 or nz = 1 states can be attributed to their even-wave character, as
discussed in the text. (c) |f |2 versus magnetic field, for a typical in-tube scattering energy of 10 kHz. The stronger coupling
of even-wave scattering is reflected in the magnetic widths of |f |2 . (d) The q1D odd-wave continuum is characterized by a
transverse wave function Ψrel (x, z) with even spatial parity. A contour plot shows the |fo |2 versus magnetic field and collision
energy. (e, f) The q1D even-wave continua are each characterized by an asymptotic transverse wave function Ψrel (x, z) with
odd spatial parity. The contour plot of |fe |2 versus magnetic field and collision energy combines the trends of panels (b) and
(c), and emphasize the growing magnetic width in the k → 0 limit that is characteristic of 1D s-wave collisions. Scattering
amplitudes in panels (b) to (f) are based on Eq. (2) with parameters in Eq. (16), or Eq. (3) with parameters in Eq. (17), with
equal VLx = VLz = 40 ER .

single quantum of excitation energy along the transverse in the strongly confined direction allows p-wave enhance-
direction. If we identify the single-particle transverse os- ment of even-wave scattering of identical fermions along
cillator ground state and first excited state as pseudo-spin the q1D direction.
down and pseudo-spin up, then this asymptotic state is We can explicitly confirm the low-energy odd-wave
a spin singlet. Just as s-wave collisions in fermionic spin scattering behaves like a true 1D system from Eq. (12)
mixtures satisfy exchange symmetry with spin-singlet by taking the low-energy limit i.e. 0 < E0 ≡ (Erel −
wave functions, here atomic wave functions entangled in ω⊥ )/2ω⊥  1:
the orbital degree of freedom allow even-wave interac-
tions along the q1D direction. A related phenomenon
 
2µ 1
occurs in s-wave scattering of non-identical particles in To−1 (Erel ) = 2
+ k ro + ik , with
2 `o
mixed dimensions [64]. In that case, the odd parity of
a2⊥ 1
 
the center-of-mass motion can induce p-wave scattering 1 2 2
= + − ζ(−1/2) ,
in the relative coordinate because the two atoms see dif- `o 6 V R3D a2⊥ a⊥
ferent trapping potentials. In our case, the CM motion a2⊥ a⊥
decouples; instead, the odd parity of the relative motion ro = + ζ(1/2) . (16)
6R3D 4
6

where ζ(s) ≡ ζ(s, 1) is the Riemann zeta function. metric trapping potentials. In our experimental system,
Equations (12) and (16) are also consistent with pre- however, the interactions are anisotropic: the 3D xy
vious calculations of p-wave scattering in q1D systems dimer branch is split by ∼ 100 kHz from the 3D z dimer
[52, 56, 58, 61, 65–67]. These equations define the effec- branch due to dipole-dipole interactions [68], leading to
tive 1D parameters for Eq. (2). direction-dependent scattering volumes V i and effective
i
Pivotal to the discussion of universality is whether ranges R3D , where i = x, y, or z [see Fig. 1(a) for axis ori-
the effective range term, kro , can be neglected. Since entation]. Furthermore the xz anisotropy of confinement
−1
ro scales as a2⊥ /R3D , it is proportional to ω⊥ , and is can be tuned by adjusting each standing-wave intensity
thus decreased with stronger confinement. In our ex- independently [as depicted in Fig. 2(a)].
periments, however, a typical kF ro ∼ 1, and thus the For broken cylindrical symmetry, N and m are no
effective-range contribution must be kept. The relatively longer good quantum numbers; we instead use cartesian
large magnitude is due to the ratio between the opti- nx and nz to index quanta of relative motion in the con-
cal confinement length scale and van der Waals length: fined directions [69], and write the states as |nx , nz , kirel .
λL /R3D is ∼ 300. (See further discussion in Sec. VI). We Even parity in y occurs for odd nx + nz , and vice versa.
also note that since the q1D approach would break down The low-dimensional scattering in each continuum is
for a⊥ . R3D , our treatment is only valid for ro  a⊥ . facilitated by the corresponding 3D p-wave dimer [see
For atoms in the lowest even-wave scattering contin- Tab. I and Fig. 2(a)]. For example, the low-dimensional
uum, i.e. N = 0 and m = ±1 or equivalently 0 < E1 ≡ interactions in the x direction are due to the coupling
(Erel − 2ω⊥ )/2ω⊥  1, the even-wave T -matrix in the to the x dimer, which has short-range symmetry corre-
low-energy limit reduces to sponding to the spherical harmonic Y1x ∼ x/r. As a
  result, the x dimer can only be hybridized with contin-
2µ i
Te−1 (Erel ) = −ae + k 2 ve + , with uum states in the |1x , 0z , kirel band above the threshold
2 k energy Ezp +~ωx . The magnetic field at which this cross-
a⊥ a3⊥
 
2a⊥ ing occurs is labeled B0x in Fig. 2(a).
ae = − + + ζ(−1/2) + ζ(1/2) ,
2 6V 3R3D Similar analyses apply for the y and z scattering and
a 3

4a⊥
 are summarized in Tab. I. The three continua have vec-
ve = ⊥ + ζ(1/2) + ζ(3/2) . (17) torially distinct coupling, with magnetically distinct res-
16 3R3D onances shown in Fig. 2(c): in order, B0y (“zeroth”), B0x
This result is identical to a true 1D calculation with s- (“first”), and B0z (“second”), and illustrated in Figs. 2(d),
wave interactions, with the effective range ve included. 2(e), and 2(f). All other combinations are forbidden by
Compared to the odd-wave effective range term, we see symmetry: i.e., the x dimer will have no effect on scat-
that to leading order k 3 ve is smaller than kro by a factor tering states |0x , 0z , kirel or |0x , 1z , kirel .
of k 2 a2⊥ , which must be small in the q1D limit. At the The q1D scattering parameters in an anisotropic trap
largest lattice depths in our experiments, kF3 ve ∼ 0.1, are still given quantitatively by Eqs. (2), (3), (16), and
which opens the possibility of exploring universal physics, (17) using the experimentally determined values of V i
i
i.e., where the effective range is negligible. and R3D [11]. This is supported by the observed data
The resonance condition for q1D scattering is `−1 o =0 collapse in Sec. IV and Sec. V. The low-energy scattering
for odd-wave or ae = 0 for even-wave. In the low-energy parameters in Eq. (16) and Eq. (17) can then be tuned
regime Em  1, the bound-state energies for the odd- by changing the 3D scattering parameters in order to en-
wave and even-wave sectors are given by gineer arbitrarily strong odd- or even-wave interactions.
In terms of the 3D quasi-bound-state energy [11, 68],
κ2o 1
EB,o = ω⊥ − with κo = − κ2o ro and (18) approximately −~2 R3D i
/(2µV i ), the resonance condition
2µ `o `−1 = 0 for odd-wave scattering in isotropic traps is
o
κ2 1
EB,e = 2ω⊥ − e with = ae + κ2e ve , (19) y
~2 R3D y
 
2µ κe R3D
− = Ezp 1 − 6 ζ(−1/2) , (20)
2µV y a⊥
where κo > 0 and κe > 0 are the odd-wave and even-
wave bound-state wave numbers, respectively. The bind- while for even-wave scattering, the resonant conditions
ing energies are consistent with true-1D calculations in
the shallow-dimer limit: for repulsive odd waves `o > 0,
ω⊥ − EB,o ≈ 1/(2µ`2o ), and for attractive even waves continuum asymptotic transverse q1D resonant
ae > 0, 2ω⊥ − EB,e ≈ 1/(2µa2e ), both in the limit state parity parity dimer
κa⊥  1. zeroth |0x , 0z , kirel even odd y dimer
first |1x , 0z , kirel odd even x dimer
second |0x , 1z , kirel odd even z dimer
B. Anisotropic interactions
TABLE I. Exchange parity and spatial orientation of the two-
The results of the previous subsection were derived atom states in the lowest three scattering continua. See also
in the limit of isotropic interactions and radially sym- Fig. 2.
7

(axe = 0 or aze = 0) are prior experimental studies in 3D for both s-wave and p-
wave systems [38, 71–75] and in quasi-two-dimensional
x,z
~2 R3D s-wave systems [76, 77]; here, we outline how both even-
− =(Ezp + ~ωx,z ) wave and odd-wave contacts can be observed in our q1D
2µV x,z
 x,z
3 R3D
 system.
× 1+ [ζ(1/2) + ζ(−1/2)] . (21)
2 a⊥
A. Loading and measurement protocol
Since R3D /a⊥ is typically small in our experiments, these
relations show that resonant odd- and even-wave scat-
tering approximately occurs when the 3D quasi-bound- An ensemble of q1D Fermi gases is prepared in sev-
state reaches the relevant scattering continuum. This eral steps. A bulk sample of 40 K in its absolute ground
behavior of the resonance condition depends only the state |ai is confined in an optical trap, and cooled sym-
energetic narrowness of the 3D p-wave Feshbach reso- pathetically with bosonic 87 Rb. At the end of the cooling
nance, and not on the geometry of the trap. This is cycle, 7(1) × 104 spin-polarized 40 K atoms at T ∼ 300 nK
self-consistent with applying Eqs. (16) and (17) to the are loaded into a lattice of tube-like anisotropic traps
case of anisotropic traps and interactions. [see Fig. 1(a)]. The confinement along the x and z di-
rections is created by two orthogonal standing waves,
Figures 2(b)–(f) show the scattering probabilities |fo |2
produced with retro-reflected beams of light with wave-
and |fe |2 , as a function of both energy and magnetic field.
length λL = 760.6 nm. Each trap is harmonic near its
A key feature is the enhanced width, in either param-
minimum, with longitudinal oscillation frequency ωy ≈
eter, of |fe |2 . Near resonance, the energetic width of
the scattering is determined by the effective range term
of Eqs. (2)–(3). For odd-wave interactions, the width
is characterized by ∆o ≡ (~2 /2µ)ro−2 , while for even- 3 State preparation Spectroscopy
Lattice potential 2
−2/3 AM
wave interactions the width is ∆e ≡ (~2 /2µ)ve

rf amplitude
. From 1 c
Eqs. (16)–(17), we find that ∆o /∆e = 18 (R3D /a⊥ )4/3 ,
2/3 Ne b b
which is small in experiments. This behavior is illus- Ng a
0
trated in Fig. 2(b) on resonance, which is a vertical cut Position t hold
at the resonant magnetic field in the contour plots of
|fo,e |2 panels (d) through (f). (e)
(a) (c)
The magnetic width of low-energy resonant scattering t AM
is instead a consequence of the scattering amplitude in (b) (d) (f)
Lattice depth

the zero-range limit. From Eq. (2), at constant `o , as


k → 0, we see that |fo |2 → 0 and scattering is thus pro- 40
0.6 40 0.6

gressively narrower in B at lower energy for odd waves 60 0.5 60 0.5

Nc
tramp 80
0.4
80
0.4

[Fig. 2(d)]. A similar analysis for Eq. (3), however, 100


0.3
100
0.3

shows fe → −1 at low energy meaning scattering be-


120

140
y
0.2
120

140
0.2
Nb
0.1

X
0.1

comes broader in B for even waves [Figs. 2(e) and 2(f)] at 160
160

Time
0
0 80 100 120 140 160 180 200 220 240 260
80 100 120 140 160 180 200 220 240 260

lower energy. At a constant scattering energy, both odd-


and even-wave resonances become broader in field with
increasing confinement. However, the odd-wave width FIG. 3. Experimental sequence. (a) Energetic depth of the
scales with the square root of the lattice depth while the confining two-dimensional optical lattice versus time. Fol-
even-wave field width scales linearly with lattice depth. lowing the creation of a degenerate Fermi gas, the optical
This can be seen from Eqs. (16)–(17) since the 3D scat- standing waves are ramped on to generate the array of q1D
traps. The two-part ramp includes a slow increase (in tramp )
tering volume V ∼ 1/B near its Feshbach resonance, so
to minimize heating followed by a fast jump to 40 ER to iso-
1/`o ∼ a2⊥ B and ae ∼ a4⊥ B [70]. In the even-wave case, late q1D ensembles. (b) At this stage, time-of-flight band
therefore, confinement can lead to a broad resonance fea- mapping typically indicates that the majority of atoms are
ture that extends the parameter regime of universality, in the ground band (central dashed box) but some atoms are
although the energetic width of the underlying 3D reso- also loaded into the first excited band (lateral boxes). (c)
nance is still narrow. The lattice depth is optionally modulated for tAM before be-
ing ramped to its full depth. The frequency is chosen to drive
excited-band atoms resonantly into the third excited band,
from which they leave the trap. (d) Time-of-flight images
III. SPECTROSCOPY
confirm that AM reduces population in the excited band. (e)
The weakly interacting gas is quenched into an interacting
Interactions between atoms in a dilute gas lead to state through a spin flip from |ai to |bi, held for thold , and
short-range correlations, which we probe via rf spec- probed with a spectroscopic pulse at detuning ωe . (f) The frac-
troscopy. These correlations are characterized by the tion of atoms transferred from |bi to |ci by the spectroscopic
contact parameters [24–40] and have been the object of rf pulse is measured with state-selective imaging.
8

2π × 400 Hz, and transverse confinement set by the lat- hyperfine-Zeeman state, adiabatically connected to the
tice depth VL , measured in units of the recoil energy, |F = 9/2, mF = −7/2i state at low magnetic field. After
ER = h2 /(2mλ2L ) ≈ h × 8.6 kHz. For equal VL = 40 ER , choosing the magnetic field B and lattice depth VL for a
for instance, ω⊥ ≈ 2π×100 kHz. This gives an anisotropy particular measurement, interactions are snapped on by
of λ ≡ ωy /ω⊥ ≈ 4 × 10−3 . The transformation of the 3D transferring the atoms to the state |bi using a 40-µs-long
cloud into an array of q1D clouds is performed in two rf pulse [see Fig. 3(e)]. After a hold time thold , typically
steps. First, the lattice depth is slowly ramped to a few 160 µs, a spectroscopic pulse of duration τ = 120 µs is
ER in tramp , typically 300 ms. Then, the lattice depth is applied. This probe has a Blackman envelope to mini-
quickly increased to 40(2) ER to form isolated tubes, see mize Fourier broadening, and a frequency that is offset ω
Fig. 3(a). We use an adiabatic loading model described above the single-particle |bi-to-|ci resonance, where |ci is
in App. A to estimate the final number and tempera- the third-lowest hyperfine-Zeeman state (and assumed to
ture distribution in the tubes. For typical experimental be weakly interacting: see App. D). The number of atoms
parameters, the peak number per tube is N ≈ 30 (thus transferred, Nc , is assessed by state-sensitive imaging in
N λ ≈ 0.1) and T /TF ≈ 0.8, where TF is the Fermi tem- time-of-flight, shown in Fig. 3(f). The measured transfer
perature. rate Γ is given by Nc /τ times a corrective factor related
Figure 3(b) shows a typical band-mapped time-of- to the pulse shape.
flight image after this loading procedure. We adjust
the chemical potential of the gas by tuning the optical
trap depth to optimize for either ground-band loading or
multi-band loading. For the latter, when the chemical
potential of the bulk gas is comparable to ER , a num- B. rf spectral form in the q1D regime
ber of atoms Ne are loaded into the first excited state of
both transverse directions since loading at the band edge
cannot be adiabatic. The fraction of atoms loaded into At high frequency, the rf spectra is related to the short-
the first excited state is typically 15% in each direction. ranged correlations in the many-body system. To quan-
To prepare for dedicated experiments measuring odd- tify this relation for both q1D odd-wave and q1D even-
wave correlations in the ground band (Sec. IV), we de- wave interactions, we derive the analytic expressions for
plete the excited-band population using amplitude mod- the rf transition rate Γ(ω) at high frequency, taking into
ulation (AM). The power of both optical standing waves account the multiple transverse excited bands. Our start-
is modulated for a time tAM at a frequency resonant with ing point is the relation between Γ and the 3D correlation
the transition from the single-atom first-to-third excited strength [27, 78]:
states. Due to the anharmonicity of the sinusoidal con-
finement, this modulation is off-resonant from the zero-
Ω2
Z Z Z
i(ω+iδ)t
to-two excitation, leaving the number of atoms in the Γ(ω) = Im i dt e d R d3 r
3

lowest single-particle state Ng relatively unaffected [see 4


    
Fig. 3(c)]. When optimizing for ground-band loading, 1 1
Tt ψb† ψc R + r, t ψc† ψb R − r, 0 ,
band mapping after AM [Fig. 3(d)] indicates that the 2 2
excited-band population is reduced to . 5% in each lat- (22)
tice direction. Lattice depth calibration is performed via
AM of the single-particle zero-to-two excitation, and we
estimate the uncertainty of the peak lattice depth to be where Ω is the Rabi frequency, r and R are the rela-
±3%. tive and COM positions, Tt is the time-ordering opera-
Ensembles for even-wave correlation measurements tor, and ψ(b,c) are the fermionic annihilation operator for
(Sec. V) are prepared either with the lattice ramped on a fermion in hyperfine-Zeeman state |bi or |ci. (Note that
without AM, or with phase modulation of the lattice ~ and m are suppressed here.) From standard arguments,
to enhance population in the first excited bands. The at high frequencies, the dominant contribution to the rf
phase modulation is implemented experimentally by a transfer rate is determined by the short-range part of the
small frequency modulation (FM) of the optical stand- many-body wave function, which is essentially the same
ing waves, which creates a time-dependent displacement as the two-body case at zero energy apart from a nor-
of local potential minima where atoms are trapped, due malization constant (which we will introduce as γ0 or γ1
to the optical propagation to the retro-reflecting mirror momentarily) that depends on the many-body physics.
and back. This effective phase modulation provides the Since we have evaluated the two-body T -matrix ex-
necessary parity-changing gradient to drive transitions actly for cylindrically symmetric q1D scattering in
from the ground to the excited single-particle band. Sec. II A, we can evaluate Eq. (22) exactly at the two-
With either preparation, the q1D ensembles of atoms body level [27, 78]. This gives us the exact functional
in the lowest hyperfine-Zeeman state |ai are long-lived, dependencies of Γ(ω) in the high-frequency limit, apart
but weakly interacting. A p-wave Feshbach resonance ex- from overall constants that characterize the many-body
ists in the bb channel [4, 5], where |bi is the second-lowest states. We find that the rf transfer rate can be written
9

2.0
P+1
as Γ(ω) = m=−1 Γ(m) (ω) with
0.12
199.2G
Ω2 (2ω⊥ )3/2

Transfer Γ (10 )
–3
Γ(0) (ω) = γ0 (23) 1.5 203.0G 0.08
8π ω2 ~ –3/2
∞   1/2   ω
X ω ω 0.04

~
× −N θ −N 1.0
2ω⊥ 2ω⊥
N =0
0.00
and 4 6 8 10 12
0.5
2 3/2
Ω (2ω⊥ )
Γ(±1) (ω) = γ1 (24)
8π ω2 0.0
∞   −1/2  
X ω ω 0 5 10
× (N + 1) −N θ −N .
2ω⊥ 2ω⊥ ~
Detuning ω
N =0

Equations (23) and (24) are valid for large detuning FIG. 4. rf spectrum of odd-wave interactions. The normal-
ω  EF and their physical contents are as follows. For ized rf transfer rate Γe versus dimensionless detuning ω e from
both Γ(0) (ω) and Γ(±1) (ω), the sum over N represents the single-particle resonance [see Fig. 3(e)]. Two spectra are
the contributions from all radially excited states. The shown, both for VL = 40ER : at B = 203.0 G (squares),
step function θ(ω/2ω⊥ − N ) indicates the opening of the gas is weakly interacting; at B = 199.2 G (circles), odd-
a new transverse band when the rf photon energy ex- wave interactions are Feshbach-enhanced [−1/kF `o ∼ 5; see
ceeds the energy difference to the N -th scattering contin- Fig. 1(b)]. The interacting gas shows an increased Γ e for posi-
uum, thus making the final scattering states within that tive ωe . The inset shows Γ(e
e e −3/2
ω ) on the same axes, with a ω
band available. This gives rise to a characteristic cusp fit to the high-frequency tail of the interacting spectrum, al-
(ω/2ω⊥ − N )1/2 for the odd-wave rf transfer rate Γ(0) , lowing the determination of C` from Eq. (29).
and the singular dependence (ω/2ω⊥ − N )−1/2 for the
even wave contribution to the rf transfer rate Γ(±1) (ω).
This feature is robust and is a consequence of the scat- Similarly, for even-wave scattering, the leading asymp-
tering amplitudes in Eq. (15) and the 1D-like density of totic spin-flip rate with characteristic 1/ω 5/2 -dependence
states of the new scattering continuum. The discontin- is a measure of the even-wave 1D contact Ca :
uous feature is observed experimentally in the m = ±1
sector (Sec. V B). Ω2 1 Ca
lim Γ(±1) (ω) = √ . (30)
Our calculation also clarifies the condition for the ex- EF /~ω<2ω⊥ 2 ~m ω 5/2
ploration of 1D physics as far as rf spectra is concerned.
In this case, the 1D regime for odd- and even-wave in- In the limit ω → ∞, increasingly shorter length scales are
teractions is achieved when ω < 2ω⊥ for which only the probed as the continuum final states of higher transverse
continuum states with N = 0 are relevant. In this case bands are made available, eventually approaching the 3D
limit; this behavior is also present in Eqs. (23),(24). Re-
2ω⊥ 1
Γ(0) (ω) = γ0 Ω2 , (25) placing the sum over transverse radial states with an in-
8π ω 3/2 tegral, one reproduces the correct leading order ω −1/2 rf
(2ω⊥ )2 1 spectral law for a p-wave Fermi gas in 3D. A related dis-
Γ(±1) (ω) = γ1 Ω2 . (26)
8π ω 5/2 cussion of the q1D momentum distribution is given by
These two relations should be identified with the expres- Ref. 80.
sions for Γ(ω) from the exact 1D odd-wave [42] and even- The anisotropy discussed in Sec. II B is observed in
wave scattering calculations: the measured contacts. The odd-wave contact is only
governed by the interactions along the y-direction, so
Ω2 Ω2 Eq. (29) is not affected. On the other hand, the even-
Γ(0) (ω) = C` and Γ(±1) (ω) = Ca , (27)
2ω 3/2 2ω 5/2 wave interactions now appear as x and z correlation
where C` and Ca are the 1D odd-wave and even-wave peaks at distinct magnetic field values, instead of the
contacts, respectively. We thus identify γ0 and γ1 as degenerate m = ±1 correlation peaks in Eq. (27). The rf
transfer rates are now proportional to Cax and Caz .
4π 4π
γ0 = C` and γ1 = Ca . (28)
2ω⊥ (2ω⊥ )2
IV. OBSERVATION OF ODD-WAVE
Restoring units, the leading asymptotic spin-flip rate CORRELATIONS
with characteristic ω −3/2 dependence is a measure of the
odd-wave 1D contact [41–46, 79] C` :
We probe odd-wave interactions by performing rf spec-
Ω2 m1/2 C` troscopy at magnetic fields near B0y . Figure 4 shows
lim (0)
Γ (ω) = + O(ω −5/2 ) . (29) e ≡ EF Γ/(~πΩ2 N )
EF /~ω<2ω⊥ 2 ~3/2 ω 3/2 the dimensionless transfer fraction Γ
10

2.5 Γ(±1)  Γ(0) , this allows a single-shot determination,


(a)
2.0 (b)
Contact C l /N (10 –2 )
2.0 40 ER C
e` √
e 3/2 Γ
= 4 2π ω e, (31)
60 ER 1.0 N
1.5 80 ER
100 ER 0.0 where
~

1.0 120 ER 198 200 202


Magnetic Field (G) e` ≡ 4m C`
C (32)
160 ER ~2 k F
0.5
is the dimensionless odd-wave contact [81]. In
0.0 Figs. 5(a,c), both loss and correlations shift to higher
magnetic field with increasing lattice depth, primarily
1.1
(c) due to the increasing zero-point energy associated with
1.0 the transverse harmonic confinement. When comparing
Atom Number (arb.)

Fig. 5(a) to Fig. 5(b), or comparing Fig. 5(c) to Fig. 5(d),


0.9 we see that field-dependent data collapses to a single
0.8 trend when plotted as a function of −1/kF `o , indicat-
40 ER
ing that we are probing the physics in the q1D regime.
0.7 60 ER 1.0
First consider the weakly repulsive regime. No signifi-
80 ER 0.8
0.6 cant correlations are seen in the range −20 < −1/kF `o <
100 ER 0.6 −2, which can be understood from Fig. 2(a): the 3D
120 ER
0.5 0.4 (d) quasi-bound p-wave dimer has yet to emerge above the
160 ER
198 200 202 scattering threshold. Similar to the 3D p-wave case
0.4 Magnetic Field (G)
[38], we conclude that the dimer state is not signifi-
–20 –10 0 10 20 30 40
cantly populated. With stronger apparent repulsion [82],
Scattering length –1/kF lo −5 . −1/kF `o < 0, losses are observed, perhaps due to
the instability of the repulsive odd-wave gas [83].
FIG. 5. Odd-wave contact and loss. (a) Contact per atom Let us next consider the regime with relatively weak
versus odd-wave scattering length, −1/kF `o . Per Eq. (31), attractive interactions, 4 . −1/kF `o . 15. In the ab-
the dimensionless odd-wave contact per atom is related to sence of strong loss, we can compare the measured values
the scaled transfer from spectroscopy. The zero-temperature of C` to calculations. Figure 5(a) shows data compared
Bethe ansatz prediction for kF ro = 1 is shown in dashed-
to zero-temperature Bethe ansatz theory [45] and to a
black, and the high-temperature prediction for T /TF = 0.8
is shown as a band covering the range of experimental lattice
high-temperature expansion theory [46] (App. B). The
depths. (b) The inset displays the same C e` /N data versus theory band for the high-temperature expansion reflects
magnetic field, emphasizing the data collapse when instead the weak 1D effective-range dependence on lattice depth
plotting versus −1/kF `o . (c) The total atom number after and includes a correction for the inhomogeneous density
the spectroscopy sequence, as a fraction of initial atom num- and tube number discussed in App. A. In this weakly
ber, is shown as a function of −1/kF `o . In the grey-shaded interacting regime, the measurements overlap both the
parameter range, 30% or more atoms are lost. Peak loss oc- zero-temperature and the high-temperature predictions,
curs at the calculated `−1
o = 0. (d) The inset displays this with no adjustable parameters, and follow the anticipated
atom-number data as a function of magnetic field. monotonic decrease with `−1 o . Note that the order of
magnitude of C e` /N is 10−2 .
At even higher magnetic fields, where −1/kF `o & 15,
the apparent contact deviates from these predictions. We
attribute this departure to overlapping even-wave corre-
versus dimensionless rf detuning ω e ≡ ~ω/EF , both
lations (see Sec. V), due to a remaining population in the
rescaled by the calculated Fermi energy of the central
excited bands, such that Γ can no longer be associated
tube (App. A). Correlations are revealed by the asym-
purely with Γ(0) .
metric spectrum at large |e ω | and by the characteristic
Finally, consider magnetic fields close to resonance
e −3/2 power-law decrease in the range 4 . ω
ω e . 15, for
(B ∼ B0y ), where we observe a smaller contact than is an-
which ω < 2ω⊥ . We fit the data in this regime to ex-
ticipated by theory. The unitary high-temperature value
tract C` , thereby measuring the odd-wave correlations in
a q1D ultracold gas for the first time to our knowledge. of the contact is C e` /N = 8/π in a uniform system [46]
while at zero temperature C e` /N = 4/π [42, 44, 79]. This
Figure 5 shows the contact strength and atom number far exceeds the highest C e` /N observed. Instead, within
measured across a range of magnetic fields and lattice the range −2 . −1/kF `o . 4, indicated in Fig. 5(c)
depths. The contact was measured using transfer fraction with a gray band, the contact per atom is decreased.
at a single frequency, typically ω
e = 5, which fulfills the This is attributed to the onset of strong loss: we find
condition EF /~  ω < 2ω⊥ of Eq. (29). Assuming that that more than 30% of the initial population is lost by
11

the end of thold , from which we infer a loss coefficient 1.0 (a) B 1st B 2nd Multi-band
L3 ≥ 10−5 cm2 /s. As in prior studies of atom loss in q1D

Transfer Γ (10 )
loading

–4
Fermi gases [22, 23], we find that L3 is independent of 0.8 y
B0 Ground-band
loading
lattice depth [see data collapse in Fig. 5(c)]. 0.6

~
Limited by losses, the system does not develop a con-
tact parameter larger than C e` /N ∼ 2 × 10−2 . We can 0.4
interpret this magnitude by estimating the odd-wave in- 0.2 B 0th
teraction energy that would be found in an equilibrium
cloud with the same correlation strength. The adiabatic 0.0
relation for 1D odd waves is [42] 198 199 200 201 202
∂F 2
~ kF e Magnetic field B (G)
= C` , (33)
∂(−`−1 ) 4m x
2.0 (b)
o

Correlation peak shift ∆B (G)


VL
where F is the free energy of the gas. Integrating
under the observed contact outside of the loss domi-
y
nated regime, we find that between (−kF `o )−1 = 10 1.5 B 0th – B 0
y
and (−kF `o )−1 = 4, the interaction energy is ∆F ∼ B 1st – B 0
−0.03 N EF . Such a q1D gas is weakly interacting. In y
B 2nd – B 0
contrast, the dimensionless contact in a 3D p-wave gas 1.0
was found to be of order unity, and a similar analysis via
the adiabatic theorem finds ∆F ∼ −0.5 N EF [38].
0.5

V. OBSERVATION OF EVEN-WAVE
CORRELATIONS 0.0
0 20 40 60 80 100
z
VL (ER)
Having understood the odd-wave correlations that are
shared between Fig. 1(b) and Fig. 1(c), we now turn
our attention to the anomalous features that appear FIG. 6. Vector assignment of the second even-wave corre-
only in Fig. 1(c). We undertake a sequence of exper- lation peak. (a) Normalized rf transfer rate Γ e at constant
iments to test the hypothesis that these spectral fea- e ≈ 5 versus magnetic field B applied in the z direction.
ω
tures are due to q1D even-wave correlations, and eluci- Three peaks in Γ(B)e are resolved for multi-band loading
date their nature: spectroscopy with variable asymmet- (circles), while a single clear peak is observed for ground-
ric confinement (Sec. V A); spectroscopy at variable fre- band loading (squares). Here, confinement is isotropic, with
quency (Sec. V B); and spectroscopy with variable cou- VLx = VLz = 40 ER . The magnetic-field displacements from
pling strength (Sec. V C). The sample preparation com- resonance position B0y are indicated for maximum Γ e are in-
mon to all these tests does not use AM cleaning (see dicated by colored arrows, and reported as ∆B in the sub-
sequent panel. (b) The magnetic-field displacements for the
Sec. III and App. C), such that ∼ 15% of the atoms are
in each of the transverse excited bands. three peaks in Γ(B)
e are shown for variable VLz , with fixed
x
VL = 40 ER . While B0th and B1st shift with the zero-point
energy included in B0y , B2nd has a stronger dependence on VLz .
This supports the hypothesis that the third resolved peak in
A. Vector character of the even-wave resonances
Γ
e versus B corresponds to scattering the even-wave z contin-
uum, whose energy is ~ωz above the odd-wave continuum.
The resonance condition for even-wave interactions is
given by Eq. (21), which can be approximated for the
i = x or i = z continuua as
from Eq. (20). In the following, we demonstrate this
~2 R3D
i

− ≈ Ezp + ~ωi . (34) anisotropic dependence of B0i on ωi .
2µV i i
B=B0 Figure 6(a) shows Γ(B)
e at VL = 40ER , a depth at
where the l.h.s. is the 3D p-wave dimer energy, linear in which three distinct peaks can be resolved in multi-band
B, as shown in Fig. 2(a). The incorporation of an ex- loading. We label the locations ofPthe peak Γ e as B0th ,
i
tra vibrational excitation quantum (shared between the B1st , and B2nd , however since Γ = Γ , additional tests
atoms in a |1x irel or |1z irel motional state) shifts the even- are necessary to identify the origin of these peaks. The
wave resonances from the odd-wave resonance condition, first peak also appears with a ground-band loading proce-
which is dure [as shown in Fig. 6(a) and Fig. 1] so it can be unam-
y biguously associated with odd-wave interactions. Note
~2 R3D that B0th is ∼ 0.2 G above B0y (vertical dashed line) due
− ≈ Ezp , (35)
2µV y
B=B0 y to loss, as explored in Fig. 5 for a wide range of VL .
12

Cylindrical symmetry is broken in our experiments by


the applied magnetic field direction (z) orthogonal to the 10.0
q1D direction (y). The field aligns the magnetic dipole
moments of 40 K, and thus splits the Feshbach resonances 5.0

Transfer Γ (10 )
–5
for the two even-wave features (see Sec. II B). This 0.5 G
splitting is already evident between B1st and B2nd in
6.0
Fig. 6(a).
1.0 –5/2
A vector assignment of these two higher peaks can be 4.0 ω
made using anisotropic confinement, VLx 6= VLz , whose an- 0.5
ticipated effect on B0x and B0z is illustrated in Fig. 2(a) 2.0

for ωz > ωx . Figure 6(b) shows the best-fit peak loca- 0.0
0.8 1.0 1.2
tions B0th , B1st , and B2nd from repeated Γ(B)
e scans at
z x 0.1
variable VL , with fixed VL = 40 ER . To subtract off 0.2 0.5 1.0 2.0
the shift in the zero-point contribution common to all Detuning ω/ω┴
three threshold energies, the displacements from B0y are
shown [84]. We see how ∆B1st is insensitive to VLz (simi-
FIG. 7. Even-wave rf spectrum. Dimensionless rf transfer
larly to ∆B0th ), while ∆B2nd has a stronger dependence
rate Γ
e versus detuning ω, for a multi-band-loaded ensemble
on VLz . Based on this trend, we can conclude that the at 200.5 G, with VL = 80 ER . Between 0.7 ω⊥ and 1.7 ω⊥ ,
second anomalous peak must be associated with confine- the signature even-wave scaling Γ e ∝ ω −5/2 is observed [see
ment along the z direction, and thus with the even-wave Eq. (27)]. A peak near ω ≈ 2ω⊥ is a signature of 1D even
threshold B0z . waves, as described in the main text. The solid lines show
A similar experimental study with variable VLx and Eq. (36), with a free vertical scale factor. The inset shows
fixed VLz indicates that the peak at B1st is associated Γ(ω)
e on a linear vertical scale in the ω −5/2 regime; a single-
with confinement along x. In sum, the vector character parameter fit determines the sum of the even-wave contacts.
and dependence on the anisotropic lateral confinement For these conditions, kF aze ≈ 0.4, compared to kF axe ≈ 1.6.
confirms the assignment of the correlation peak at B1st
in Fig. 6 to the x even-wave continuum, and at B2nd to
the z even-wave continuum. frequencies, below 0.7 ω⊥ /2π ∼ 90 kHz. This may be
due to incomplete thermalization of the long-wavelength
degrees of freedom.
B. Spectroscopic signatures A sharp peak in Γ(ω)
e appears at the anticipated lo-
cation of the singularity in Eq. (36), where ω ≈ 2ω⊥ ,
The strongest evidence for the observation of even- twice the band gap [85]. Here, atom pairs are promoted
wave correlations in our system comes from the func- from incident states with principal radial quantum num-
tional form of the rf spectra. As discussed in Sec. III B, ber N = 0 to final states with N = 1 with hyperfine-
within the range of frequencies ω < 2ω⊥ , in which a Zeeman state |ci, while conserving parity (i.e., conserving
q1D system resembles a true 1D system, the rf transfer m as discussed in Sec. II A). Since rf transitions cannot
rate should scale as ω −5/2 (as opposed to ω −3/2 found modify the spatial wave functions, the presence of the fea-
for odd-wave correlations in Sec. IV). Furthermore, for ture demonstrates that the scattering states are dressed
larger rf frequencies, Γ(ω)
e should exhibit a singularity. by higher-N orbital states at short range, r . a⊥ [see
Specifically, for 0 < ω < 4ω⊥ , Eq. (24) reduces to Eq. (14)]. The cusp feature in theory results from a com-
bination of two factors: the divergence of the 1D density
Ω2 Cai θ(ω)
 
i 2θ(ω − 2ω⊥ ) of final states and the constant even-wave matrix element
Γ (ω) = √ + , (36) [see Eq. (8)]. As observed, the 2ω⊥ cusp is softened by
2 ~m ω 5/2 (ω − 2ω⊥ )1/2 ω 2
the distributions of a finite range of collision energies, the
where Cai is the 1D even-wave contact for i = x or i = z. average over variable lattice depths of 1D tubes, and the
The experimentally measured Γ(ω) is a sum of all Γi , but Fourier broadening of the rf pulse.
can be dominated by only one i near its resonance. The observation of these two even-wave signatures in
Figure 7 shows the observed Γ(ω)
e spectrum at B ≈ Γ(ω) gives strong evidence for the emergence of effective
B1st . For frequencies between 0.7 ω⊥ and 1.7 ω⊥ , we ob- s-wave correlations in the q1D gas.
serve a clear −5/2 power law consistent with 1D s-wave
correlations (see Fig. 7 inset). In this region, it is pos-
sible to extract the 1D even wave contact parameters C. Even-wave contact parameter
Cax,z . We note that once Cax and Caz are fixed, the gen-
eral form of Γ(ω)
e [Eq. (24) and in particular Eq. (36)] is The correlations appearing in Fig. 1(c) across a broad
then determined for higher frequencies, no matter what range in magnetic field can now be quantified through
the relative strengths of Cax and Caz . We do not under- the even-wave contact. At each magnetic field, the even-
stand why the −5/2 power law does not extend to lower wave contacts Cax,z can be probed with a single-shot
13

x
measurement of Γ at a fixed detuning with the range Scattering length kF ae
0.7 ω⊥ . ω . 1.7 ω⊥ , as discussed in Sec. V B. In the 2 3 4 5 6 7
regime where Γ(0)  Γ(±1) , 0.8
3 5 7

Atom no. (arb.)


eax + C
C eaz = 2 2πN ωe 5/2 Γ
e, (37) 1.0

0.6
where 0.8

Contact Ca /Ne
eax,z ≡ (4m/~2 )Cax,z /kF3
C (38) 0.4 0.6
0 2 4 6

is the dimensionless even-wave contact. Since only ∼ 15%


of atoms populate each transverse excited band, we base
0.2
kF on the total atom number. We show in App. C that rf
transfer is proportional to the excited-band population,
motivating us to report the (extensive) contact divided 80 ER
0.0
by the number of atoms in one single-particle excited 120 ER
motional excited band Ne , which we assume to be equal
for both the x and z directions. In Fig. 8 we report Cea /Ne 0 1 2 3 4 5 6
for various interaction strengths, tuning the scattering z
Scattering length kF ae
length aze across a range of 200 G to 205 G, away from
the odd-wave feature. We observe data collapse between
FIG. 8. Even-wave correlations. Even-wave contact C ea =
different lattice depths and single-shot detunings ω ∼ ω⊥ Ceax + C
eaz normalized by Ne , plotted versus even-wave scat-
chosen to be within the −5/2 power law regime described tering lengths aze (bottom) and axe (top). The scattering
in Sec. V B. lengths ae are tuned by the magnetic field across a range
For kF aze & 2.5, the observed trend matches that of a 200 G to 205 G, and scaled by the calculated kF of the cen-
high-temperature model (App. B) that maps the imbal- tral tube. Single-shot measurements of Ca use Eq. (37), re-
anced band mixture in our system to a spin mixture with peated here for two different single-shot detunings: ω/2π =
true 1D s-wave interactions. For the large lattice depths 100 kHz (squares) and ω/2π = 125 kHz (diamonds). Two
in Fig. 8, the distinct resonances overlap [see Fig. 1(c)]: lattice depths are shown: red and purple points are 80 ER
the slightly weaker axe is shown on the upper axis of Fig. 8. and 120 ER , respectively. The shaded band gives a high-
temperature expansion prediction for C eax /Ne + C
eaz /Ne , from
Both data and theory include both x and z even-wave in-
teractions. However in the region of interest, correlations Eq. (B5), with no fit parameters. The width of the band
reflects one standard deviation of systematic uncertainties
are dominated by even-wave z pairing: a solid line shows
(atom number and band-population fractions) and the vari-
Ceaz /Ne alone.
ation of effective-range contributions. Both data and theory
For 0 ≤ kF aze . 2, the observed C ea /Ne plateaus to include both x and z even-wave interactions. However in the
0.5(1). The expected near-unitary contact can be esti- region of interest, correlations are dominated by even-wave z
mated using a zero-temperature theory [41] developed pairing: a solid line shows C eaz /Ne alone. The inset shows the
for 1D s-wave spin mixtures, again using a mapping of relative atom number after spectroscopy, for the same ranges
imbalanced band populations to imbalanced spin popu- of kF aze (bottom) and kF axe (top).
lation in the model. In the homogeneous case, one finds
Ceaz /Ne ≈ 1.0, on the same order of magnitude as the
observed value. A full description of the unitary regime
would need to include loss [see inset of Fig. 8], inhomo- ∆F ∼ −0.7 Ne EF . The q1D even-wave gas is thus
geneity, and the contribution of Cax . The contrast with strongly interacting, with an attractive energy that is
the q1D odd-wave case, where the observed contact was at least order of magnitude larger than is observed near
. 10−2 of the unitary value, is already striking. the odd-wave resonance.
We can interpret the strength of the observed contact
using the adiabatic relation from 1D s-waves [74]: One can understand the enhanced strength of q1D
even-wave interactions through the proportionality of p-
∂F ~2 kF3 e z wave coupling to momentum, as indicated by Eq. (8).
= C . (39) For odd waves, the relevant momentum that leads to a
∂aze 4m a
finite interaction is in the y direction, along the tube.
Assuming thermal equilibrium, the interaction energy is For even waves, p-wave coupling is achieved through the
then found by integrating under the observed contact, transverse gradient in the spatial wave function, which is
which contains contributions from both x and z reso- proportional to a−1 ⊥ [see Eq. (9)]. The square of these mo-
nances but is dominated by C eaz . As a lower estimate mentum scales appears in the matrix element Eq. (9); for
for the total interaction energy, we find that between a typical momentum kF , the ratio of odd- to even-wave
kF aze = 5 and kF aze = 2, ∆F ∼ −0.2 Ne EF , whereas coupling is a2⊥ kF2 /2, which is of order N λ and necessarily
integrating between kF aze = 5 and kF aze = 0 we find small in the q1D regime.
14

VI. DISCUSSION AND CONCLUSION especially the low-frequency deviations from the high-
frequency power law. The sub-leading contacts, conju-
The observation of emergent s-wave interactions has gate to ro and ve for the odd- and even-wave cases [46],
important implications for the prospects of universality should appear as additional power-law contributions to
in spin-polarized Fermi gases. The scaling of elastic scat- Γ.
tering near the Feshbach resonance helps illustrate this A shallow q1D dimer state lies below threshold both
point. As discussed in Sec. I, if ae is tuned to zero [for in- in the repulsive `o < 0 gas and in the attractive ae > 0
stance at B = B0z defined by Eq. (21)], then the unitary gas probed in our experiment. Searching for these dimer
limit fe → −1 is achieved in the limit ve k 3  1. The states and characterizing their stability would be inter-
range of scattering energies in which this occurs is given esting. In the even-wave case, the binding energy is large
by ~2 k 2 /2µ . ∆e , where ∆e ≡ ~2 ve
−2/3
/2µ. In terms of enough near resonance to invalidate the q1D treatment.
lateral confinement strength, the leading-order width is Another prospect is to search for emergent s-waves
[86] in quasi-two-dimensional (q2D) ensembles of identical
fermions. A similar mechanism would apply: that an-
√ tisymmetric orbital wave functions in a strong confine-
∆e ≈ ~ω⊥ (3 2R3D /a⊥ )2/3 . (40)
ment direction would enable exchange-even collisional
Within an degenerate Fermi gas, the range of two-body wave functions in the low-energy q2D degrees of freedom.
collision energies is 2EF ; these fall within the unitary In either q1D or q2D, a study of the universal dynamics
energetic width when ∆e & 2 EF , or and thermodynamics of the system would be interesting.
New universal contact relations should be found to
N λ . (3R3D /2a⊥ )2/3 . (41) clarify, for instance, the correspondence to the closed-
channel fraction. Qualitatively, a q1D (or q2D) polar-
The r.h.s. of Eq. (41) is ∼ 0.2 for the confinement ized gas with emergent s-wave interactions is expected to
strength typically used in low-dimensional quantum gas be similar to a spin mixture with direct s-wave interac-
experiments, and thus poses a comparable constraint to tions. However for systems such as metastable helium or
the q1D criterion N λ  1. For the N λ ∼ 0.1 condi- those with strong spin-orbit relaxation, spin polarization
tions in our experiments, resonant even-wave scattering is necessary for collisional stability. For fermionic iso-
is already energetically broad [see Fig. 2(b)]. topes in these systems, emergent s-wave interactions in
Elastic odd-wave collisions also have a zero-range limit. the quasi-low-dimensional regime could provide a path-
A parallel argument finds that when `−1 o → 0, the en- way to low-energy universality.
ergetic width of the unitary regime (∆o = ~2 /2µro2 ) is In summary, our observations and analysis reveal the
2
∆o ≈ ~ω⊥ (18R3D /a2⊥ ) to leading order [87]. Compar- first complete picture of quasi-one-dimensional fermions
ing this width to 2 EF , one finds that the confinement interacting with short-range p-wave interactions. The
condition for ∆o & 2 EF is interplay of exchange symmetry and orbital degrees of
freedom gives rise to a surprising feature: effective s-
N λ . (3R3D /a⊥ )2 . (42) wave interactions along the q1D direction, whose low-
energy behavior is qualitatively different from all p-wave
The r.h.s. of Eq. (42) is ∼ 0.02 for our typical conditions,
systems studied to date. In particular, unlike 3D p-waves
so is more restrictive than the q1D criterion, and not
or q1D odd-wave scattering in the ground band, even-
satisfied for our experiments [see Fig. 2(b)].
wave scattering is energetically broad.
A separate constraint is posed by three-body loss.
These energetic widths ∆o,e assume that `−1  kF and
ae  1/kF for a collision at the Fermi energy. As the den- ACKNOWLEDGMENTS
sity and energy of system is reduced, this becomes a pro-
gressively more demanding constraint for odd waves, but
The authors would like to thank S. Smale for his early
less demanding for even waves. This is depicted numeri-
contributions to the experiment, Xiwen Guan and Xiang-
cally by the widths of the unitary (red) range of magnetic
guo Yin for discussion of Bethe ansatz calculations, and
fields at low collision energy in Fig. 2(d) and Fig. 2(f).
D. Clément for discussions. This research is supported
In practice, entering the `o  ro regime in which ∆o
by NSERC, AFOSR FA9550-19-1-7044 and FA9550-19-
is defined was associated by strong three-body loss, as
1-0365, ARO W911NF-15-1-0603, HK GRF 17304820,
shown in Fig. 5(b). This mechanism was has been stud-
17304719 and CRF C6009-20G and C7012-21G.
ied for odd-wave scattering both experimentally [22, 23]
and theoretically [44]. A similar study of losses and band
relaxation in the even-wave case is called for. A possibil-
ity is that the Pauli suppression of three-body recombi-
nation in spin-singlet mixtures might also help stabilize
orbital singlet mixtures.
Future work could also include a more detailed under-
standing of rf spectra such as shown in Fig. 4 and Fig. 7,
15

[1] C. Chin, R. Grimm, P. S. Julienne, and E. Tiesinga, Fes- [19] D. V. Kurlov and G. V. Shlyapnikov, Two-body relax-
hbach resonances in ultracold gases, Rev. Mod. Phys. 82, ation of spin-polarized fermions in reduced dimensionali-
1225 (2010). ties near a p-wave Feshbach resonance, Phys. Rev. A 95,
[2] W. Zwerger, The BCS-BEC Crossover and the Unitary 032710 (2017).
Fermi Gas (Springer, Berlin, 2012). [20] L. Zhou and X. Cui, Stretching p-wave molecules by
[3] I. Bloch, J. Dalibard, and W. Zwerger, Many-body transverse confinements, Phys. Rev. A 96, 030701(R)
physics with ultracold gases, Reviews of Modern Physics (2017).
80, 885 (2008). [21] F. Fonta and K. M. O’Hara, Experimental conditions
[4] C. A. Regal, C. Ticknor, J. L. Bohn, and D. S. Jin, for obtaining halo p-wave dimers in quasi-one-dimension,
Tuning p-wave interactions in an ultracold Fermi gas of Phys. Rev. A 102, 043319 (2020).
atoms, Phys. Rev. Lett. 90, 053201 (2003). [22] Y.-T. Chang, R. Senaratne, D. Cavazos-Cavazos, and
[5] J. Zhang, E. G. M. van Kempen, T. Bourdel, R. G. Hulet, Collisional loss of one-dimensional fermions
L. Khaykovich, J. Cubizolles, F. Chevy, M. Teichmann, near a p-wave Feshbach resonance, Phys. Rev. Lett. 125,
L. Tarruell, S. J. J. M. F. Kokkelmans, and C. Salomon, 263402 (2020).
p-wave Feshbach resonances of ultracold 6 Li, Phys. Rev. [23] A. S. Marcum, F. R. Fonta, A. M. Ismail, and K. M.
A 70, 030702(R) (2004). O’Hara, Suppression of three-body loss near a p-wave res-
[6] C. H. Schunck, M. W. Zwierlein, C. A. Stan, S. M. F. onance due to quasi-1D confinement, arXiv:2007.15783
Raupach, W. Ketterle, A. Simoni, E. Tiesinga, C. J. (2020).
Williams, and P. S. Julienne, Feshbach resonances in [24] S. Tan, Energetics of a strongly correlated Fermi gas,
fermionic 6 Li, Phys. Rev. A 71, 045601 (2005). Ann. Phys. 323, 2952 (2008).
[7] F. Chevy, E. G. M. van Kempen, T. Bourdel, J. Zhang, [25] S. Tan, Large momentum part of a strongly correlated
L. Khaykovich, M. Teichmann, L. Tarruell, S. J. J. M. F. Fermi gas, Ann. Phys. 323, 2971 (2008).
Kokkelmans, and C. Salomon, Resonant scattering prop- [26] S. Tan, Generalized virial theorem and pressure relation
erties close to a p-wave Feshbach resonance, Phys. Rev. for a strongly correlated Fermi gas, Ann. Phys. 323, 2987
A 71, 062710 (2005). (2008).
[8] J. P. Gaebler, J. T. Stewart, J. L. Bohn, and D. S. Jin, [27] E. Braaten and L. Platter, Exact relations for a strongly
p-wave Feshbach molecules, Phys. Rev. Lett. 98, 200403 interacting Fermi gas from the operator product expan-
(2007). sion, Phys. Rev. Lett. 100, 205301 (2008).
[9] Y. Inada, M. Horikoshi, S. Nakajima, M. Kuwata- [28] E. Braaten, D. Kang, and L. Platter, Universal relations
Gonokami, M. Ueda, and T. Mukaiyama, Collisional for a strongly interacting Fermi gas near a Feshbach res-
properties of p-wave Feshbach molecules, Phys. Rev. onance, Phys. Rev. A 78, 053606 (2008).
Lett. 101, 100401 (2008). [29] S. Zhang and A. J. Leggett, Sum-rule analysis of radio-
[10] J. Fuchs, C. Ticknor, P. Dyke, G. Veeravalli, E. Kuhnle, frequency spectroscopy of ultracold Fermi gas, Phys. Rev.
W. Rowlands, P. Hannaford, and C. J. Vale, Binding A 77, 033614 (2008).
energies of 6 Li p-wave Feshbach molecules, Phys. Rev. [30] F. Werner, L. Tarruell, and Y. Castin, Number of closed-
A 77, 053616 (2008). channel molecules in the BEC-BCS crossover, Euro.
[11] D. J. M. Ahmed-Braun, K. G. Jackson, S. Smale, Phys. J. B 68, 401 (2009).
C. J. Dale, B. A. Olsen, S. J. J. M. F. Kokkelmans, [31] S. Zhang and A. J. Leggett, Universal properties of the
P. S. Julienne, and J. H. Thywissen, Probing open- and ultracold Fermi gas, Phys. Rev. A 79, 023601 (2009).
closed-channel p-wave resonances, Phys. Rev. Research [32] E. Braaten, D. Kang, and L. Platter, Short-time operator
3, 033269 (2021). product expansion for rf spectroscopy of a strongly inter-
[12] C. A. Bertulani, H.-W. Hammer, and U. van Kolck, Ef- acting Fermi gas, Phys. Rev. Lett. 104, 223004 (2010).
fective field theory for halo nuclei: shallow p-wave states, [33] E. Braaten, Universal Relations for Fermions with Large
Nuc. Phys. A 712, 37 (2002). Scattering Length, in The BCS-BEC Crossover and the
[13] P. Bedaque, H.-W. Hammer, and U. van Kolck, Narrow Unitary Fermi Gas, edited by W. Zwerger (Springer,
resonances in effective field theory, Phys. Lett. B 569, Berlin, 2012) pp. 193–231.
159 (2003). [34] F. Werner and Y. Castin, General relations for quantum
[14] K. B. Gubbels and H. T. C. Stoof, Theory for p-wave gases in two and three dimensions: I. Two-component
Feshbach molecules, Phys. Rev. Lett. 99, 190406 (2007). fermions, Phys. Rev. A 86, 013626 (2012).
[15] E. Braaten, P. Hagen, H. W. Hammer, and L. Plat- [35] F. Werner and Y. Castin, General relations for quantum
ter, Renormalization in the three-body problem with gases in two and three dimensions. II. Bosons and mix-
resonant p-wave interactions, Phys. Rev. A 86, 012711 tures, Phys. Rev. A 86, 053633 (2012).
(2012). [36] S. M. Yoshida and M. Ueda, Universal high-momentum
[16] H. W. Hammer and D. Lee, Causality and universality in asymptote and thermodynamic relations in a spinless
low-energy quantum scattering, Phys. Lett. B 681, 500 Fermi gas with a resonant p-wave interaction, Phys. Rev.
(2009). Lett. 115, 135303 (2015).
[17] H. W. Hammer and D. Lee, Causality and the effective [37] Z. Yu, J. H. Thywissen, and S. Zhang, Universal Re-
range expansion, Ann. Phys. 325, 2212 (2010). lations for a Fermi Gas Close to a p-Wave Interaction
[18] J. Levinsen, N. R. Cooper, and V. Gurarie, Stability of Resonance, Phys. Rev. Lett. 115, 135304 (2015).
fermionic gases close to a p-wave Feshbach resonance, [38] C. Luciuk, S. Trotzky, S. Smale, Z. Yu, S. Zhang, and
Phys. Rev. A 78, 063616 (2008). J. H. Thywissen, Evidence for universal relations describ-
16

ing a gas with p-wave interactions, Nature Phys. 12, 599 and V. Gritsev, Exact Solution for 1D Spin-Polarized
(2016). Fermions with Resonant Interactions, Phys. Rev. Lett.
[39] M. He, S. Zhang, H. M. Chan, and Q. Zhou, Concept of a 104, 040402 (2010).
contact spectrum and its applications in atomic quantum [58] K. Kanjilal and D. Blume, Nondivergent pseudopoten-
Hall states, Phys. Rev. Lett. 116, 045301 (2016). tial treatment of spin-polarized fermions under one- and
[40] S.-G. Peng, X.-J. Liu, and H. Hu, Large-momentum dis- three-dimensional harmonic confinement, Phys. Rev. A
tribution of a polarized Fermi gas and p-wave contacts, 70, 042709 (2004).
Phys. Rev. A 94, 063651 (2016). [59] L. Pricoupenko, Modeling interactions for resonant p-
[41] W.-B. He, Y.-Y. Chen, S. Zhang, and X.-W. Guan, Uni- wave scattering, Phys. Rev. Lett. 96, 050401 (2006).
versal properties of Fermi gases in one dimension, Phys. [60] Z. Idziaszek, Analytical solutions for two atoms in a har-
Rev. A 94, 031604(R) (2016). monic trap: p-wave interactions, Phys. Rev. A 79, 062701
[42] X. Cui, Universal one-dimensional atomic gases near odd- (2009).
wave resonance, Phys. Rev. A 94, 043636 (2016). [61] S. Saeidian, V. S. Melezhik, and P. Schmelcher, Shifts
[43] X. Cui and H. Dong, High-momentum distribution with and widths of p-wave confinement induced resonances in
a sub-leading k−3 tail in odd-wave interacting one- atomic waveguides, J. Phys. B 48, 155301 (2015).
dimensional Fermi gases, Phys. Rev. A 94, 063650 [62] B. E. Granger and D. Blume, Tuning the interactions of
(2016). spin-polarized fermions using quasi-one-dimensional con-
[44] Y. Sekino, S. Tan, and Y. Nishida, Comparative study finement, Phys. Rev. Lett. 92, 133202 (2004).
of one-dimensional Bose and Fermi gases with contact [63] G. Moore, T. Bergeman, and M. Olshanii, Scattering in
interactions from the viewpoint of universal relations for tight atom waveguides, J. Phys. IV 116, 69 (2004).
correlation functions, Phys. Rev. A 97, 013621 (2018). [64] Y. Nishida and S. Tan, Confinement-induced p-wave reso-
[45] X. Yin, X.-W. Guan, Y. Zhang, H. Su, and S. Zhang, nances from s-wave interactions, Phys. Rev. A 82, 062713
Momentum distribution and contacts of one-dimensional (2010).
spinless Fermi gases with an attractive p-wave interac- [65] D. Muth, M. Fleischhauer, and B. Schmidt, Discretized
tion, Phys. Rev. A 98, 023605 (2018). versus continuous models of p-wave interacting fermions
[46] J. Maki, Thermodynamic contacts and breathing-mode in one dimension, Phys. Rev. A 82, 013602 (2010).
physics of one-dimensional p-wave Fermi gases in the [66] B. Heß, P. Giannakeas, and P. Schmelcher, Energy-
high-temperature limit, Phys. Rev. A 104, 063314 dependent `-wave confinement-induced resonances, Phys.
(2021). Rev. A 89, 052716 (2014).
[47] D. Blume and C. H. Greene, Fermi pseudopotential ap- [67] K. Jachymski, F. Meinert, H. Veksler, P. S. Julienne, and
proximation: Two particles under external confinement, S. Fishman, Ultracold atoms in quasi-one-dimensional
Phys. Rev. A 65, 043613 (2002). traps: A step beyond the Lieb-Liniger model, Phys. Rev.
[48] P. Naidon, E. Tiesinga, W. F. Mitchell, and P. S. Juli- A 95, 052703 (2017).
enne, Effective-range description of a Bose gas under [68] C. Ticknor, C. A. Regal, D. S. Jin, and J. L. Bohn, Mul-
strong one- or two-dimensional confinement, New J. tiplet structure of Feshbach resonances in nonzero partial
Phys. 9, 19 (2007). waves, Phys. Rev. A 69, 042712 (2004).
[49] T. Kristensen and L. Pricoupenko, Ultracold-atom col- [69] The cylindrically symmetric states are superpositions of
lisions in atomic waveguides: A two-channel analysis, cartesian states with nx + nz = 2N + |m|.
Phys. Rev. A 91, 042703 (2015). [70] The widening field range of correlations at higher lattice
[50] K. Günter, T. Stöferle, H. Moritz, M. Köhl, and depths can be seen in Fig. 1, with the effect strongest
T. Esslinger, p-wave interactions in low-dimensional for even-wave correlations in Fig. 1(c). There, even-wave
fermionic gases, Phys. Rev. Lett. 95, 230401 (2005). correlations extend across a 3 G range at VL = 120 ER ,
[51] M. Waseem, Z. Zhang, J. Yoshida, K. Hattori, T. Saito, compared to a . 0.5 G range of odd-wave correlations at
and T. Mukaiyama, Creation of p-wave Feshbach similar lattice depth. Both of these are larger than the
molecules in selected angular momentum states using an ∼ 0.2 G range of significant correlations observed in 3D.
optical lattice, J. Phys. B 49, 204001 (2016). [71] W. Schneider and M. Randeria, Universal short-distance
[52] T.-Y. Gao, S.-G. Peng, and K. Jiang, Two-body state structure of the single-particle spectral function of dilute
with p-wave interaction in a one-dimensional waveguide Fermi gases, Phys. Rev. A 81, 021601(R) (2010).
under transversely anisotropic confinement, Phys. Rev. [72] J. T. Stewart, J. P. Gaebler, T. E. Drake, and D. S. Jin,
A 91, 043622 (2015). Verification of universal relations in a strongly interacting
[53] M. Olshanii, Atomic scattering in the presence of an ex- Fermi gas, Phys. Rev. Lett. 104, 235301 (2010).
ternal confinement and a gas of impenetrable bosons, [73] E. D. Kuhnle, H. Hu, X.-J. Liu, P. Dyke, M. Mark, P. D.
Phys. Rev. Lett. 81, 938 (1998). Drummond, P. Hannaford, and C. J. Vale, Universal be-
[54] J. I. Kim, J. Schmiedmayer, and P. Schmelcher, Quan- havior of pair correlations in a strongly interacting Fermi
tum scattering in quasi-one-dimensional cylindrical con- gas, Phys. Rev. Lett. 105, 070402 (2010).
finement, Phys. Rev. A 72, 042711 (2005). [74] M. Barth and W. Zwerger, Tan relations in one dimen-
[55] V. Dunjko, M. G. Moore, T. Bergeman, and M. Olshanii, sion, Ann. Phys. 326, 2544 (2011).
Confinement-induced resonances, in Advances in Atomic, [75] S. Hoinka, M. Lingham, K. Fenech, H. Hu, C. J. Vale,
Molecular, and Optical Physics, Vol. 60, edited by E. Ari- J. E. Drut, and S. Gandolfi, Precise determination of the
mondo, P. Berman, and C. Lin (Academic Press, 2011) structure factor and contact in a unitary Fermi gas, Phys.
pp. 461–510. Rev. Lett. 110, 055305 (2013).
[56] L. Pricoupenko, Resonant scattering of ultracold atoms [76] B. Fröhlich, M. Feld, E. Vogt, M. Koschorreck, M. Köhl,
in low dimensions, Phys. Rev. Lett. 100, 170404 (2008). C. Berthod, and T. Giamarchi, Two-dimensional Fermi
[57] A. Imambekov, A. A. Lukyanov, L. I. Glazman, liquid with attractive interactions, Phys. Rev. Lett. 109,
17

130403 (2012). Appendix A: Inhomogeneous averaging


[77] C. Luciuk, S. Smale, F. Böttcher, H. Sharum, B. A.
Olsen, S. Trotzky, T. Enss, and J. H. Thywissen, Obser-
The theoretical description of atom-atom correlations
vation of quantum-limited spin transport in strongly in-
teracting two-dimensional Fermi gases, Phys. Rev. Lett.
in Secs. II and III assume a single tube with a homoge-
118, 130405 (2017). neous atomic density. Our experimental measurements
[78] D. T. Son and E. G. Thompson, Short-distance and average across two types of inhomogeneity: the density
short-time structure of a unitary Fermi gas, Phys. Rev. profile within a single tube, and tube-to-tube variation
A 81, 063634 (2010). in N . In the high-temperature limit, Chom /N n1D for
[79] S. A. Bender, K. D. Erker, and B. E. Granger, Exponen- a uniform-density gas is solely a function of tempera-
tially decaying correlations in a gas of strongly interact- ture, scattering parameters, and confinement strength
ing spin-polarized 1D fermions with zero-range interac- (see Sec. B), i.e., variables that are uniform across our
tions, Phys. Rev. Lett. 95, 230404 (2005). ensemble. We can thus account for inhomogeneities by
[80] M. He and Q. Zhou, p-wave contacts of quantum gases in separating out the dependence of contact on atom num-
quasi-one-dimensional and quasi-two-dimensional traps,
ber and density, as
Phys. Rev. A 104, 043303 (2021).
[81] Such that the momentum density nk → C̃` kF /k2 , when Chom
R
(dk/2π)nk = N . hCiobs = hN n1D i (A1)
N n1D
[82] Note that inhomogeneous transverse confinement broad-
ens the range of `−1 o for a given magnetic field, which where hN ni is an average across the ensemble.
makes the assignment of small −1/kF `o ambiguous. The inhomogeneity within each tube is treated with
[83] L. Pan, S. Chen, and X. Cui, Many-body stabilization a local density approximation: that each differential
of a resonant p-wave Fermi gas in one dimension, Phys. length dy contributes n21D (y)dy to N n1D . In the high-
Rev. A 98, 011603 (2018). temperature limit, the density distribution is given by a
[84] The sub-leading term of the isotropic Eq. 20 depends on Maxwell-Boltzmann distribution, so
a single a⊥ , but provides only a ∼ 50 mG correction to Z
the leading order. Still, we determine B0y through a gen-
hN n1D itube = n21D (y)dy = (4π)−1/2 N 2 /RT (A2)
eralized form of Eq. (20) that includes anisotropic con-
finement [52].
[85] Due to anharmonicity of the optical lattice potential, where RT = (kB T /mωy2 )1/2 is the thermal length of
there is some ambiguity in the experimental value of the the cloud. By comparison, the peak 1D density √ is
transverse trap frequency. Since the lowest allowed exci- (2π)−1/2 N/RT , such that hN n1D itube = N npk / 2.
tation for even-waves is the excitation of an atom from We model the distribution of q1D ensembles by assum-
the first to third single-particle level of the optical lattice, ing the transfer between a 3D Fermi gas and an array of
the best choice for ~ω⊥ in this case is half that energy. tubes is isentropic. This approach neglects the small frac-
[86] The sub-leading correction modifies ∆e by 1 −
tion of atoms that appear in higher bands, since edges
2R3D a−1⊥ [ζ(1/2) + ζ(3/2)], such that the leading order
gives roughly a 10% underestimate for our typical pa-
of the Brillouin zone cannot be adiabatic. This trans-
rameters. fer occurs during the first tramp phase of Fig. 3(a), up
[87] The sub-leading correction modifies ∆o by 1 − to a modest depth (VL < 10ER ). The density of states
3R3D ζ(1/2)/a⊥ , such that the leading order gives of the low-depth array is a convolution
√ of a√harmonic-
roughly a 20% underestimate for our typical parameters. lattice density of states ρ() = 2 ER /π~ω⊥  [88, 89]
[88] A. M. Rey, G. Pupillo, C. W. Clark, and C. J. Williams, in the x and z directions with a 1D harmonic oscillator
Ultracold atoms confined in an optical lattice plus ρ() = 1/~ωy along the tube. We find the chemical po-
parabolic potential: A closed-form approach, Phys. Rev. tential and temperature of the transformed tube array by
A 72, 033616 (2005). imposing entropy and number conservation between the
[89] P. B. Blakie, A. Bezett, and P. Buonsante, Degener- pre-loaded gas to the array of tubes. After this loading
ate Fermi gas in a combined harmonic-lattice potential,
phase, energy and particle transport between the tubes
Phys. Rev. A 75, 063609 (2007).
[90] C. Gao and Z. Yu, Breathing mode of two-dimensional are suppressed as the lattice depths increase to their final
atomic Fermi gases in harmonic traps, Phys. Rev. A 86, values. Including a −20% corrective factor for atom loss
043609 (2012). observed in round-trip loading calibrations, we find the
[91] J. Hofmann, Quantum anomaly, universal relations, and number and temperature the central tube values to be
breathing mode of a two-dimensional Fermi gas, Phys. N ∗ ≈ 30 and Te ≈ 0.8 for typical conditions; here and
Rev. Lett. 108, 185303 (2012). below, we will use N ∗ to define a reference value of EF .
[92] Y.-C. Zhang and S. Zhang, Strongly interacting p-wave The atom-number distribution of this adiabatic loading
Fermi gas in two dimensions: Universal relations and model is compared to a diabatic model in which the ini-
breathing mode, Phys. Rev. A 95, 023603 (2017).
tial 3D density distribution is partitioned into an square
[93] H. Hu and X.-J. Liu, Resonantly interacting p-wave
Fermi superfluid in two dimensions: Tan’s contact and
array of parallelepipeds. In both models, the distribution
the breathing mode, Phys. Rev. A 100, 023611 (2019). of N is found to be roughly uniform: i.e., as many tubes
have near-peak atom number as have only one atom.
We define aPdistribution function wN of atom number
that satisfies N wN = M , where M is the total number
18

where ΛT = (2π~2 β/m)1/2 is the single-particle thermal


P
of tubes, and N wN N = Ntot . Without a dispersion in
atom number, each of the M tubes would have N ∗ atoms, de Broglie wavelength. As assumed in App. A, the con-
so that Ntot = N ∗ M , and the distribution function would tact is N n times a function of T , `o , and ro . A convenient
be wN = Ntot δN,N ∗ /N ∗ . Instead, in the distribution form uses kF and EF of the central tube as reference val-
exp
found in loading models, wN is approximately constant ues:
between N = 1 and N = N ∗ , and zero otherwise. This
~2 kF3 Te`2o
 
gives C` 3 2
= 1 + kF ro `o T ,
e (B2)
N 2πm 2

N
exp
X 1 ∗ exp where we have used Λ−2 e 2
Ntot = wN N= (N + 1)N ∗ wN . (A3) T = T kF /4π and kF = πn.
2
N =1

exp 2. Even wave


We can rearrange this expression to find wN =
∗ ∗ ∗
2Ntot /[(N + 1)N ] for 1 ≤ N ≤ N . Since the con-
tact is proportional to N 2 in the high-temperature limit The calculation technique for the even-wave contact
[see Eq. (A2)], the ratio of the contacts in these two cases follows the high-temperature treatment for odd waves
is given by [46], i.e., using a two-channel model of spin-half fermions

coupled to closed-channel molecules to evaluate the con-
N
hN 2 iexp 1 X exp 2 1 + 2N ∗ tacts for arbitrary interaction strengths. To model the
= w N N = . (A4) anisotropic nature of the interactions and the trapping
(N ∗ )2 Ntot N ∗ 3N ∗
N =1 potential, we assume that the x and z resonances are
independent 1D systems. The expression for the con-
which approaches 2/3 for large N ∗ .
tact Cax,z has a contribution from the scattering contin-
Combining the intra-tube and inter-tube inhomogene-
uum (Ca |at ) and a contribution from the two-body bound
ity effects yields a correction in the estimated value of
state (Ca |mol ). Omitting the direction label (x, z), these
the contact
two terms are
hCtube iexp
r
2 π e−1/2 Z ∞
2π −1/2 u1/2 e−u
= T . (A5)

Chom 3 8 Ca at = Ne npg
du 2 ,
0 (mu/~2 ) (ae − muve /β~2 ) + β
(B3)
where Te = kB T /EF uses EF for the central tube, where the integration variable is dimensionless, pg is the
~ωy N ∗ . The theoretical curves shown in all figures use population fraction of ground-band atoms, and
the uniform-gas prediction multiplied by this correction, √
which is ∼ 0.4 for typical loading parameters. ~2 2ΛT θ(ae ) β~2 /(ma2e )
Ca mol ≈ Ne npg
e , (B4)
m ve + |ae |3 /2

Appendix B: High-temperature expansion for the where θ(x) is the Heaviside step function. Note that in
odd- and even-wave contacts Eq. (B4), we have used the shallow-dimer limit instead
of the full Eq. (19), which is a good approximation in the
In the high-temperature regime of small fugacity, z  regime kF ae & 2 shown in Fig. 8. Just like the odd-wave
1, analytic forms of the many-body contact can be found case, the even-wave contact is a function of N n, but now
from a virial expansion [46, 90–93]. Since our loading multiplied by the fraction in each band.
model predicts weakly degenerate 1D ensembles with Scaling by kF and EF of the central tube, the sum of
Te . 1, this is an appropriate limit in which to compare these contributions is
theory and experiment. In each case, we consider a sin-  √ 2 2e
Ca 2pg ~2 kF3 2πe2/(kF ae T ) e−1/2
gle 1D system with uniform density n and total number = 3/2 T
Ne π m kF3 ve + |kF ae |3 /2
N = Ln. Inhomogeneity is accounted for as described in
Z ∞
u1/2 e−u

App. A.
+ du , (B5)
0 u(kF ae − ukF3 ve Te/2)2 + 2Te−1
1. Odd wave which is plotted in Fig. 8. In the weakly interacting limit
the even-wave contacts are given by
The high-temperature calculation of the odd-wave con- Ca 2pg ~2 kF −2
tact comes from a two-channel model of interacting ≈ ae (B6)
Ne πm
atoms and closed-channel molecules [46]. In the weakly
interacting limit `o /ΛT  1, In the resonantly interacting limit, the high tempera-
ture theory breaks down, as the binding energy ∝ 1/a2e
2π~2 N n`2o
 
`o ro quickly exceeds the temperature, invalidating the high-
C` = 1 − 6π , (B1)
mΛ2T Λ2T temperature assumption.
19

(a) (c) (d) (e)


3
2.2

Lattice potential
2
AM 1.2
1 2.0
Ne

Relative transfer (arb.)


Ng 1.8
1.0
0 1.5
Position
1.4
(b) 0.8
3
Lattice potential

FM 2 1.0
1 1.0
0.6
0.0 0.5 1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
0 AM strength (arb.) FM strength (arb.) Population product
Position 3
Odd-wave peak Even-wave (x) peak Ne Ng /N (10 )

FIG. 9. Correlation of transfer with modulation type and strength. (a) Amplitude modulation of the lattice potential causes
a single-particle first-to-third transition that removes atoms from the first excited band. (b) Frequency modulation causes
a single-particle zero-to-first transition to populate the excited band while also depleting the ground band. (c) Number of
transferred atoms to probe state from ω e /2π = 50 kHz for 40 ER confinement with varying AM strength. No significant change
in transfer at the odd-wave peak (199.25 G). Less transfer is noticed at the even-wave (x) peak (199.75 G). Transfer is plotted
relative to the transfer with no modulation. (d) Transferred atoms for various frequency modulation. A slight decrease in
transfer for the odd-wave peak is seen as atoms are transferred out of the ground band. Transfer is increased for the even-wave
peak. (e) Transfer at even-wave peak compared to the band population product Ne Ng /N with a linear fit.

Appendix C: RF spectroscopy with variable band FM modulates atoms from the ground band to the ex-
populations cited band, one might expect an equal change in the
strength of correlations for odd and even-waves. In-
The odd- and even-wave contacts are extensive quan- stead, we observe that increasing modulation strength
tities that are proportional to population in the relevant only slightly decreases transfer at the odd-wave peak and
bands at high temperature. We are able to manipulate causes comparatively large increases in transfer for the
band populations with two types of lattice modulations, even-wave peak. From this we remark that the relatively
amplitude modulation (AM) and frequency modulation equal weights for the peaks seen in Fig. 1(c) is coinciden-
(FM). In this Appendix, we show their effects on the tal for the ∼ 15% of excited-band atoms populated by
measured contacts. our multi-band loading procedure.
Figure 9(a) illustrates AM and its primary effect on Figure 9(e) demonstrates how transfer at the even-
band populations. Anharmonicity of the lattice allows wave peak is correlated with the product of ground and
for the tuning of the frequency of modulation to be such excited state atom populations, counted in band mapped
that only one transition is resonant: AM frequency res- time-of-flight imaging. A linear fit of the measurements
onant first-to-third excited states transfer is off-resonant has a zero crossing within uncertainty of zero and a re-
for excitation from the ground band to the second excited duced χ2 of 1.07. This measurement provides us with the
band. Since atoms in the third band leave the trap, this confidence to attribute the transfer seen at the even-wave
process distills the ground band. Figure 9(b) illustrates peaks as a measure of correlations between ground and
how FM of the lattice produces a local position displace- excited band atoms.
ment of the lattice potential. Unlike AM, this can drive
parity-changing transitions, and transfer atoms from the
ground band to the first excited band. Appendix D: Final-state effects in even-wave
The effect of AM on the odd- and even-wave contacts spectroscopy
are shown in Fig. 9(c). Here circles and squares report
Γ at magnetic fields near the odd-wave and even-wave The ideal scenario of rf spin-flip spectroscopy is that an
peaks, respectively, which is proportional to the (exten- atom is transferred from an interacting state to a non-
sive) contact. We see that increasing AM strength has interacting state. This is a good approximation when
no noticeable trend in Γ at the odd-wave peak, which is interactions in the initial spin state are near-resonant,
consistent with odd-wave correlations being due to scat- and the final states at the same magnetic field are far
tering between ground-band atoms. At the even-wave from resonance. In the situation probed in our exper-
peak, decrease of Γ with AM is also consistent with even- iments, atoms are spin-flipped from |bi to |ci, so that
wave correlations being proportional to the exited-band bb p-wave interactions should be large compared to bc
population. p-wave interactions. Coupled-channels calculations indi-
Similarly, the effect of FM is shown in Fig. 9(d). Since cate that the z-dimer bc p-wave Feshbach resonance is
20

at B0 = 217.12 G, with a scattering volume that can of frequency and scattering parameters relevant to the
be parameterized as Vbc (B)/Vbg = 1 − ∆/(B − B0 ), with experiment, it does not explain the deviation of the even-
Vbg = (−107.95a0 )3 and ∆ = −8.98 G. This yields a zero- wave rf spectrum from the −5/2 scaling shown in Fig. 7.
crossing near 208 G, such that Vbc is below its background In our regime of interest, Eq. (D1) can be approxi-
value in the range of fields across which we measure q1D mated as
contacts, typically 200 G – 205 G.
For even-wave spectroscopy, in place of Eq. (30), the 2
relation between Γ and Ca is Ω2 Ca (ae − a0e )
lim Γ(ω) = √ . (D2)
ω→∞ 2 ~m ω 5/2 a02
e
2
Ω2 Ca (ae − a0e − mω(ve − ve0 )/~)
lim Γ(ω) = √ ,
ω→∞ 2 ~m ω 5/2 (mωve0 /~ − a0e )2 + (mω/~)−1 This allows us to bound the final-state effects to correc-
(D1) tions between 0.90 and 1.05 using Vbc and a typical value
where a0e and ve0 are the 1D scattering parameters for of R3D in the field and depth range used in Fig. 8. Since
the bc channel. Equation (D1) predicts a deviation of this is less than the scatter of the data, we can safely
Γ(ω) from its asymptotic form, however within the range neglect final-state effects.

You might also like